Influence diagnostics in the varying coefficient model with longitudinal data (Q626197)

A linear regression model with time-varying coefficients is considered for longitudinal data. A local constant estimate is used for the coefficients as functions of time. The authors propose a version of Cook's distance as an influence measure of a set of observations. In this distance a jackknife estimate for the covariance of the estimates is used. A bootstrap technique is proposed to derive reference values for these distributions. An application to AIDS clinical data is presented.